In this paper we study the stochastic batch sizing problems. We provide a unifying treatment of the problem, in which we formulate a multi-stage recourse problem as well as a probabilistically constrained problem. The solution approach that we adopt for these problems may be classified as a branch and price (B&P) method. Through our computational experiments turns out that the proposed B&P methodology is quite effective for the recourse constrained model. We also demonstrate how tradeoffs between cost and reliability can be investigated for the stochastic batch-sizing problem.